Archive for January, 2005

Quantifying the Learning Curve

Sunday, January 2nd, 2005

Back in the 1980s, R.W. Hockney introduced two measures for quantifying the performance of pipelined machines. The first, r∞, is the pipeline's maximum possible performance when startup effects are ignored (or equivalently, its performance on an infinitely-long vector). The second, n½, is the half-performance length, i.e. how much ...